A Machine Learning Framework for Estimating Atherosclerosis Severity Levels Using Cardiovascular Data

Abstract

Accurate assessment of atherosclerosis severity is critical for reducing cardiovascular morbidity and mortality. This study evaluates machine learning models for estimating atherosclerosis severity using cardiovascular risk factors from the Atherosclerosis Risk in Communities (ARIC) dataset, comprising 15,792 participants. Demographic, clinical, and laboratory features including blood pressure, cholesterol, glucose, and carotid artery intima-media thickness were used. The data were split into training (70%) and testing (30%) sets, and seven supervised models were assessed: Random Forest, Support Vector Machine, Naïve Bayes, Logistic Regression, Gradient Boosting Machine, XGBoost, and K-Nearest Neighbors. Performance was evaluated using accuracy, precision, recall, and F1 score. XGBoost achieved the highest performance with an accuracy of 0.89 and F1 score of 0.88, followed by Random Forest with an accuracy of 0.88 and F1 score of 0.87, while Naïve Bayes performed the worst with an accuracy of 0.79. Correlation analysis revealed strong associations between carotid intima-media thickness and cholesterol, as well as between systolic and diastolic blood pressure. These results demonstrate the effectiveness of ensemble-based machine learning models for estimating atherosclerosis severity and supporting clinical risk stratification.